Three AI IPOs, $3.5 Trillion — Can Public Markets Tell the Difference Between Growth and Profit?

Three companies. Three IPO filings. A combined target valuation north of $3.5 trillion. The AI economy is about to face its first real test: can these companies convince public market investors that AI is not just usage growth — but durable, compounding profits?

The Three Filings

Anthropic filed a confidential S-1 on June 1 at a $965 billion valuation. The company raised $65 billion in its Series H days earlier, tripling its February valuation. Claude is the default enterprise AI for coding and research workflows. Revenue is estimated at $8-12 billion annualized, growing triple digits. Anthropic has no consumer app with 400 million users. It has something arguably more valuable: the cleanest corporate structure, no nonprofit conversion baggage, and the enterprise market where margins are highest.

OpenAI is preparing its own confidential filing for late 2026 at approximately $852 billion. ChatGPT — as explored in the intelligence factory race between AI labs — has 400 million+ weekly active users. Revenue reportedly crossed $15 billion annualized. But OpenAI carries structural complexity: the nonprofit-to-profit conversion is still being litigated, board governance questions persist, and the Microsoft relationship — while enormously valuable — creates dependency questions that public market analysts will probe relentlessly.

SpaceX-xAI prices on June 11 at a $1.75 trillion target after merging Starlink’s infrastructure — as explored in the economics of AI compute infrastructure — with Grok AI. The roadshow launches this week. A $75 billion raise via 21 banks. Combined revenue of $18.67 billion in 2025, but a $4.94 billion net loss from merger costs. This is the wildest bet of the three: Elon Musk is arguing that AI + space infrastructure = a category that doesn’t exist yet.

The Real Question: Usage Growth vs. Durable Profits

Every AI company can show usage curves going up and to the right. The question public markets will ask is different: what are the unit economics at scale?

Anthropic and OpenAI face the same structural challenge. Their primary cost — compute — scales roughly linearly with usage. Every API call, every inference request, every agent action consumes GPU time. Unlike software companies where marginal cost approaches zero, AI companies pay for every token. The path to durable profits requires one of three things: pricing power that exceeds compute cost growth, inference efficiency improvements that bend the cost curve, or vertical integration that captures more of the value chain.

Nvidia’s Vera Rubin promising 10x cheaper inference helps the model providers — but it also helps their competitors equally. Cheaper compute is a rising tide, not a moat.

SpaceX-xAI has a different challenge entirely. Investors must believe that combining satellite internet with an AI model creates synergies that justify $1.75 trillion. The bull case — AI needs global connectivity, Starlink provides it, the combined entity becomes the infrastructure layer of the AI economy — is compelling but unproven. The bear case: these are two separate businesses duct-taped together by one person’s vision.

Can Public Markets Absorb $3.5 Trillion in AI?

The aggregate demand on public markets is staggering. SpaceX-xAI alone seeks $75 billion in fresh capital. If Anthropic and OpenAI each raise $20-40 billion in their IPOs, the total capital pulled from public markets for AI could exceed $130 billion in a single year.

That money comes from somewhere. Institutional investors reallocate from existing positions — which means selling something else to buy AI. If all three listings happen in 2026, the capital rotation out of traditional tech, healthcare, and energy into AI becomes one of the largest sector rotations in market history.

The timing matters. If SpaceX-xAI prices successfully on June 12 and trades up, it creates momentum for Anthropic and OpenAI. If it prices down or trades poorly, the entire AI IPO window could narrow — forcing Anthropic and OpenAI to delay or accept lower valuations.

The next six months will answer the biggest question in the AI economy: is this a generational technology shift that justifies generational valuations, or is this 1999 with better models? The IPO race is where theory meets capital — and capital always tells the truth eventually.

For the full structural map of the AI economy, read The Map of AI Redrawn on Business Engineer.